/LSTM_LDA

Combining LDA with LSTM network

Primary LanguagePython

Setup virtual environment

Clone the repository to your computer and create a virtual environment

$ git clone https://github.com/jjlatval/LSTM_LDA.git $ cd LSTM_LDA $ virtualenv venv $ source venv/bin/activate

Install requirements

In virtualenv run: $ pip install -r requirements.txt

How to run LSTM + LDA

  1. Adjust n_topics and n_cpu_cores to your liking in topic_modelling.py
  2. Calculate LDA beta matrix by running calculate_lda.py
  3. Adjust network parameters (vocabulary_size, hidden_dim, learning_rate & nepoch) in lstm_model.py
  4. Train network weights by running train_lstm.py
  5. Generate output text by using the trained network in lstm_output.py
  6. Run analytics on generated text by running analytics.py

Known issues

lstm_model gets stuck sometimes in the forward propagation step for an unknown reason